← CHI 2026 map

CHI '26 · Honorable mention · full-paper review · confidence medium-high

From Junior to Senior: Allocating Agency and Navigating Professional Growth in Agentic AI-Mediated Software Engineering

Dana Feng , Bhada Yun , April Yi Wang

DOI PDF Program page Page JSON

This is a strong mixed-methods CHI paper with a clear human-centered reframing: agency in AI-mediated software engineering is treated as an organizational and developmental issue, not just a usability issue. The main contribution is a practice proposal grounded in junior and senior accounts, with explicit limits that keep the claims appropriately scoped.

Axes Lens

Rare contribution shape, typical evidence profile. The point here is not a score. It is to show what kind of claim the paper makes, and whether the evidence pattern is unusual or baseline in this 268 -review set.

Review Summary

This paper stands out because it shifts the discussion of agentic AI in software engineering away from the usual productivity or code-quality lens and toward agency allocation, professional growth, and mentorship across career stages. The evidence base is a three-phase mixed-methods design that combines senior elicitation, junior task performance, and senior review of junior prompt histories, which gives the authors multiple vantage points on how AI changes control, delegation, and learning. The most compelling claim is that agency is shaped upstream by organizational policies and tooling defaults, not merely by individual preference at the moment of interaction. That is a useful corrective to common-sense narratives that treat AI use as a purely personal choice. The paper’s novel move is not a new algorithm or interface, but a synthesized practice proposal, Prompt & Code Reviews, meant to make AI use more accountable and thereby preserve agency. As a CHI contribution, that is a legitimate and timely form of novelty: normative and practice-oriented rather than technical. The limitations are also well articulated and materially important. The authors acknowledge that the study is a snapshot, that the phases used different tasks so juniors and seniors cannot be directly compared, and that the sample is small and regionally concentrated. Those constraints reduce overclaim risk and make the findings best read as a grounded organizational argument rather than a general law of AI-mediated work. Overall, this is a credible honorable-mention-level paper because it offers a coherent conceptual reframing, a plausible practice implication, and a careful account of scope.

What Changed

Canon before

Prior CHI work on AI-assisted software engineering has often emphasized productivity, code generation, or developer experience; this paper shifts the center of gravity toward agency allocation, professional growth, and mentorship across junior and senior developers.

Departure from common sense

The paper argues that agency in AI-mediated software work is not mainly a matter of personal preference or individual skill, but is preconfigured by organizational policies, tooling defaults, repos, and CI guardrails before developers act. That framing pushes against the intuitive view that agency is primarily negotiated at the moment of use.

Actual novelty

The paper’s main novel contribution is a synthesized practice proposal, Prompt & Code Reviews (PCRs), intended to restore agency through accountability by selectively documenting prompts and involving senior oversight. The novelty is less a new tool than a new accountability practice for AI-mediated software engineering.

Evidence

The paper uses a three-phase mixed-methods design: ACTA plus Delphi with 5 seniors, an AI-assisted debugging task with 10 juniors, and blind reviews of junior prompt histories by 5 additional seniors. The abstract and discussion frame the findings around agency allocation, professional growth, and mentorship, while the limitations section explicitly notes that the phases used different tasks and cannot support direct junior-vs-senior comparison.

Limits

Method limits

The study is a snapshot in time, uses different tasks across phases so juniors and seniors cannot be directly compared in this paper, and relies partly on retrospective accounts. The sample is small and regionally concentrated, which constrains inference.

Deployment limits

The proposed practices are framed for software engineering organizations that already use agentic AI and have policies, tooling defaults, and review processes capable of supporting accountability. Transfer to other domains or less structured teams may require adaptation.

Boundary conditions

Findings are bounded by short-term situational tasks, a small sample, and a USA/Canada-concentrated participant pool. The paper itself notes that differing phase structures prevent direct comparison between junior and senior groups.

Manual check: No manual check required beyond standard merge QA; evidence is grounded in the provided abstract and limitation excerpts, but the exact wording of the proposed PCR practice should be verified against the full text if publication requires stricter citation precision.

Position in field

This is a CHI contribution that reframes agentic AI in software engineering from a productivity-centered topic to one about agency, accountability, and career development. Its value lies in synthesizing cross-generational accounts into a practice-oriented argument rather than introducing a technical system.

Abstract

Juniors enter as AI‑natives, seniors adapted mid‑career. AI is not just changing how engineers code–it is reshaping who holds agency across work and professional growth. We contribute junior–senior accounts on their usage of agentic AI through a three-phase mixed-methods study: ACTA combined with a Delphi process with 5 seniors, an AI-assisted debugging task with 10 juniors, and blind reviews of junior prompt histories by 5 more seniors. We found that agency in software engineering is primarily constrained by organizational policies rather than individual preferences, with experienced developers maintaining control through detailed delegation while novices struggle between over-reliance and cautious avoidance. Seniors leverage pre-AI foundational instincts to steer modern tools and possess valuable perspectives for mentoring juniors in their early AI-encouraged career development. From synthesis of results, we suggest three practices that focus on preserving agency in software engineering for coding, learning, and mentorship, especially as AI grows increasingly autonomous.